Skip to main content

Foundational Trials

TTM

Image: Shutterstock / ame435

Publication

  • Title: Targeted Temperature Management at 33°C versus 36°C after Cardiac Arrest
  • Acronym: TTM
  • Year: 2013
  • Journal published in: New England Journal of Medicine
  • Citation: Nielsen N, Wetterslev J, Cronberg T, et al. Targeted temperature management at 33°C versus 36°C after cardiac arrest. N Engl J Med. 2013;369:2197-2206.

Context & Rationale

  • Background
    • Mild therapeutic hypothermia (typically 32–34°C) became standard after early RCTs (2002) suggested improved neurological outcomes versus then-usual care (often with fever).
    • The evidence base for an “optimal” target temperature was limited, with uncertainty around whether benefit derived from true hypothermia or simply avoidance of hyperthermia.
    • Deeper cooling carries biologically plausible harms (electrolyte shifts, arrhythmias, coagulopathy, infection risk, prolonged sedation/paralysis), creating a credible equipoise for a higher controlled target.
    • Post–cardiac arrest care was evolving (coronary angiography/PCI, haemodynamic optimisation, structured neuroprognostication), potentially altering the marginal benefit attributable to temperature alone.
  • Research Question/Hypothesis
    • In comatose adult survivors of out-of-hospital cardiac arrest of presumed cardiac cause, does targeting 33°C improve survival and neurological outcome compared with targeting 36°C, when both groups receive active temperature management and contemporary ICU post-arrest care?
  • Why This Matters
    • Clarified whether the “dose” of temperature (33°C vs higher controlled normothermia) meaningfully changes outcomes, rather than conflating benefit with fever avoidance.
    • Had immediate implications for international practice, because 33°C had become widely adopted despite limited comparative data.
    • Provided a platform to standardise neuroprognostication and reduce self-fulfilling prophecy bias in postarrest trials, which is central to credible causal inference in this population.

Design & Methods

  • Research Question: Among unconscious adults resuscitated after out-of-hospital cardiac arrest (presumed cardiac cause), does targeted temperature management at 33°C reduce all-cause mortality compared with targeted temperature management at 36°C?
  • Study Type: Randomised, international, multicentre (36 ICUs), parallel-group trial; allocation concealed; temperature strategy open-label; neuroprognostication structured and designed to minimise bias.
  • Population:
    • Setting: Intensive care units (with enrolment from ICU or emergency department) across Europe and Australia.
    • Key inclusion: Adults (≥18 years); out-of-hospital cardiac arrest of presumed cardiac cause; sustained ROSC >20 minutes; unconscious at screening (per protocol: Glasgow Coma Scale <8).
    • Key exclusions: Time from ROSC to screening >240 minutes; unwitnessed arrest with initial asystole; suspected intracranial haemorrhage or stroke; initial temperature <30°C.
  • Intervention:
    • Target: 33°C.
    • Delivery: Active temperature control using site-available cooling methods (e.g., surface or intravascular devices), with sedation/analgesia and anti-shivering measures as required.
    • Duration and rewarming: Target temperature maintained until 28 hours after randomisation; rewarming to 37°C at 0.5°C per hour; ongoing fever prevention until 72 hours after ROSC.
  • Comparison:
    • Target: 36°C.
    • Delivery: Active temperature control to maintain 36°C (including warming if needed), with comparable sedation/anti-shivering strategy to enable target adherence.
    • Duration and fever prevention: Same time-structure as intervention (28-hour maintenance window, controlled rewarming, and fever prevention to 72 hours post-ROSC).
  • Blinding: Treating teams were not blinded to target temperature; mortality is objective, but open-label care can influence co-interventions and withdrawal decisions. The protocol emphasised structured neuroprognostication to reduce differential withdrawal-of-life-sustaining-therapy bias.
  • Statistics: A total of 900 patients with complete follow-up were required to detect a hazard ratio of 0.80 for death (33°C vs 36°C) with 90% power at a two-sided 5% significance level; planned enrolment was 950 to allow for attrition. Primary analysis used a modified intention-to-treat approach with time-to-event methods (Cox regression), with additional sensitivity analyses.
  • Follow-Up Period: 180 days (mortality and neurological outcomes assessed at 180 days).

Key Results

This trial was not stopped early. Recruitment and follow-up proceeded to the planned sample size.

Outcome 33°C 36°C Effect p value / 95% CI Notes
All-cause mortality through end of trial (180 days; primary) 235/473 (50%) 225/466 (48%) HR 1.06 95% CI 0.89 to 1.28; P=0.51 Time-to-event; modified intention-to-treat.
Poor neurological outcome at 180 days (CPC 3–5) 242/469 (52%) 247/464 (53%) RR 0.98 95% CI 0.88 to 1.10; P=0.78 Risk ratio among patients with available 180-day CPC.
Poor neurological outcome at 180 days (mRS 4–6) 239/469 (51%) 247/464 (53%) RR 0.96 95% CI 0.85 to 1.08; P=0.48 Risk ratio among patients with available 180-day mRS.
Death at 180 days (secondary) 236/473 (50%) 225/466 (48%) RR 1.01 95% CI 0.87 to 1.15; P=0.92 Binary mortality at 180 days (secondary).
Any serious adverse event excluding death (days 1–7) 439/472 (93%) 420/466 (90%) Not reported P=0.086 Supplementary appendix; serious adverse events collected days 1–7.
Hypokalaemia (serious adverse event; days 1–7) 86/472 (19%) 58/466 (13%) Not reported P=0.018 Supplementary appendix; electrolyte disturbance consistent with deeper cooling physiology.
  • There was no statistically significant difference in 180-day mortality (50% at 33°C vs 48% at 36°C; HR 1.06; 95% CI 0.89 to 1.28; P=0.51) and no signal of improved neurological outcome by CPC or mRS at 33°C.
  • Serious adverse events (days 1–7) were common in both groups; hypokalaemia was more frequent with 33°C (19% vs 13%; P=0.018), while overall serious adverse events excluding death did not differ significantly (93% vs 90%; P=0.086).
  • Prespecified subgroup analyses did not demonstrate a clear effect-modification signal (e.g., time to ROSC ≤25 min: HR 0.92; 95% CI 0.68 to 1.24 vs >25 min: HR 1.20; 95% CI 0.96 to 1.50; interaction P=0.20; shock at admission absent: HR 1.03; 95% CI 0.83 to 1.28 vs present: HR 1.35; 95% CI 0.90 to 2.03; interaction P=0.17).

Internal Validity

  • Randomisation and Allocation
    • Computer-generated assignment sequences with concealed allocation and undisclosed block sizes; stratified by site.
    • Baseline characteristics were closely balanced, supporting successful randomisation.
  • Drop out or exclusions
    • Screened: 1431; randomised: 950.
    • Withdrawal of consent: 1/476 (0.2%) in 33°C; 3/474 (0.6%) in 36°C.
    • Modified intention-to-treat exclusions after randomisation: 2 in 33°C; 5 in 36°C (final modified intention-to-treat n=473 vs n=466).
    • Neurological outcomes at 180 days were available for 469/473 (33°C) and 464/466 (36°C) for CPC/mRS-derived endpoints.
  • Performance/Detection Bias
    • Treating clinicians were not blinded to temperature target, which can influence co-interventions and timing of neurological assessment.
    • Mortality is objective; neurological endpoints (CPC/mRS) remain vulnerable to bias mediated through withdrawal of life-sustaining therapy and assessment variability.
    • Withdrawal of life-sustaining therapy within 10 days was similar: HR 1.11; 95% CI 0.88 to 1.40; P=0.38; median time to withdrawal 5 days (IQR 3–8) in both groups (supplementary appendix).
  • Protocol Adherence
    • Per-protocol population: 472 (33°C) vs 464 (36°C).
    • Crossovers were rare: received wrong intervention 0 vs 1 (supplementary appendix).
    • Early rewarming in the 33°C group occurred for clinical reasons: arrhythmia (n=6), severe circulatory instability (n=4), bleeding (n=2), uncontrolled lactate rise (n=2), urgent CABG (n=1) (supplementary appendix).
  • Baseline Characteristics
    • Groups were broadly comparable in pre-arrest and arrest characteristics (e.g., witnessed arrest 93% in both groups; bystander CPR 74% vs 75%; initial shockable rhythm 359/473 vs 348/466).
    • Contemporary coronary care was frequent and balanced (coronary angiography on day 1: 282/473 [60%] vs 283/466 [61%]; PCI on day 1: 188/473 [40%] vs 182/466 [39%]).
    • Illness severity markers at admission were similar (circulatory shock 15% vs 14%; ST-elevation myocardial infarction 40% vs 42%).
  • Heterogeneity
    • International multicentre design increases heterogeneity in delivery (cooling devices, sedation practice), but site stratification and large sample size mitigate confounding.
    • No clinically persuasive subgroup effect modification was apparent on prespecified analyses.
  • Timing
    • Enrolment allowed up to 240 minutes from ROSC to screening, so temperature assignment generally occurred after hospital arrival rather than intra-arrest or immediate post-ROSC.
  • Dose
    • Both groups received protocolised temperature control and fever prevention; the “dose” difference was a 3°C target separation during the early postarrest phase.
    • Electrolyte physiology differed as expected: hypokalaemia was more frequent at 33°C (19% vs 13% among serious adverse events recorded days 1–7; P=0.018).
  • Separation of the Variable of Interest
    • Highest recorded temperature on day 2: 36.0 ± 1.5°C (33°C group) vs 37.2 ± 0.7°C (36°C group); P<0.001 (supplementary appendix).
    • Highest recorded temperature on day 3: 37.7 ± 0.5°C vs 37.8 ± 0.6°C; P=0.017 (supplementary appendix).
    • Highest recorded temperature on day 4: 37.8 ± 0.6°C vs 37.9 ± 0.7°C; P=0.002 (supplementary appendix).
    • Hours with temperature >38°C: median 0 (IQR 0–0) vs 0 (IQR 0–1); P=0.070; rebound hyperthermia >39°C: 0 vs 0 (supplementary appendix).
  • Key Delivery Aspects
    • The comparison was between two actively managed targets; the trial does not test “hypothermia vs no temperature control”.
    • ICU length of stay was similar: median 8 days (IQR 4–13) vs 8 days (IQR 4–14); hospital length of stay was similar: median 11 days (IQR 6–22) vs 10 days (IQR 5–21) (supplementary appendix).
  • Outcome Assessment
    • Primary endpoint (all-cause mortality) is robust and clinically meaningful.
    • Neurological endpoints used CPC and mRS at 180 days, which are widely used but relatively coarse; nevertheless they align with practice-relevant disability thresholds.
  • Statistical Rigor
    • Prespecified effect size and high power (90%) for a hazard ratio of 0.80; primary analysis used time-to-event modelling with modified intention-to-treat.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong given concealed randomisation, balanced baseline characteristics, high follow-up completeness, and clear temperature separation early after arrest; the main threat is open-label delivery in a context where withdrawal and co-interventions can influence neurological endpoints.

External Validity

  • Population Representativeness
    • Enrolled a predominantly witnessed out-of-hospital cardiac arrest population with high rates of bystander CPR and shockable rhythms, consistent with many EMS systems where postarrest ICU admission is routine.
    • Key exclusions (e.g., unwitnessed asystole, suspected intracranial catastrophe, very low initial temperature, prolonged ROSC-to-screening time) reduce applicability to higher-risk and non-cardiac aetiology cohorts.
  • Applicability
    • Findings are most applicable to centres able to deliver protocolised, resource-intensive temperature control (devices, sedation/neuromuscular blockade, frequent monitoring) and structured neuroprognostication.
    • Generalisation to settings with limited ICU capacity, delayed postarrest transport, or different coronary reperfusion pathways may be constrained.

Conclusion on External Validity: Generalisability is good for comatose adult OHCA survivors managed in well-resourced ICUs with contemporary postarrest care pathways, but is limited for unwitnessed asystolic arrests, non-cardiac causes, and resource-constrained systems.

Strengths & Limitations

  • Strengths:
    • Large, international, multicentre RCT in a domain historically dominated by smaller single-region trials.
    • Clinically meaningful primary endpoint with near-complete ascertainment at 180 days.
    • Both arms used active temperature management and fever prevention, isolating target temperature rather than conflating benefit with “any temperature protocol”.
    • Structured approach to postarrest care and withdrawal decisions, reducing a key bias pathway in neurocritical outcomes research.
  • Limitations:
    • Open-label delivery may influence co-interventions and withdrawal-of-life-sustaining therapy behaviour, particularly for neurological outcomes.
    • Enrolment allowed up to 4 hours post-ROSC, so the trial does not answer whether earlier cooling (including intra-arrest) modifies effect.
    • Neurological outcome measures (CPC/mRS) are widely used but relatively coarse and may miss more subtle cognitive differences.
    • Physiological adverse effects of deeper cooling were evident (e.g., hypokalaemia), and early rewarming occurred for clinical instability in some patients.

Interpretation & Why It Matters

  • What the trial shows
    • In a contemporary postarrest care context with active temperature control in both arms, targeting 33°C did not improve survival or global functional outcome compared with targeting 36°C.
  • Clinical practice implication
    • The results support prioritising meticulous temperature control (particularly fever prevention) rather than presuming that deeper hypothermia (33°C) is universally required.
    • Where 33°C is chosen, clinicians must anticipate and manage predictable physiological consequences (e.g., hypokalaemia) and instability prompting early rewarming.
  • Trialist/methodology implication
    • TTM reframed the “control” condition as an actively treated comparator and highlighted withdrawal-of-care processes and neuroprognostication as integral components of internal validity in postarrest RCTs.

Controversies & Subsequent Evidence

  • Interpretation error: “TTM proves hypothermia does not work” is not what was tested
    • The trial compared two actively enforced temperature targets (33°C vs 36°C), with fever prevention in both arms, so it does not address “hypothermia vs no temperature management”. 1
    • Misapplication of the results to justify passive temperature management risks reintroducing fever as an uncontrolled co-intervention. 1
  • Time-to-target and biological timing debate
    • Allowance of up to 240 minutes from ROSC to screening raised concern that any time-sensitive benefit of earlier cooling could be diluted, particularly if the hypothesised mechanism depends on immediate reperfusion injury modulation. 2
  • Subsequent RCT evidence and the direction of travel
    • TTM2 (NEJM 2021) compared hypothermia (33°C) with targeted normothermia (fever prevention) after out-of-hospital cardiac arrest and similarly did not show improved survival with routine hypothermia; arrhythmias requiring treatment were more frequent with hypothermia. 3
    • Collectively, these trials have shifted the evidentiary centre-of-gravity towards fever prevention and careful temperature control, with ongoing debate about whether particular biological/clinical subgroups benefit from deeper hypothermia.

Summary

  • In 939 unconscious OHCA survivors (modified intention-to-treat), targeting 33°C did not reduce 180-day all-cause mortality compared with targeting 36°C (50% vs 48%; HR 1.06; 95% CI 0.89 to 1.28; P=0.51).
  • Global neurological outcomes at 180 days were similar: CPC 3–5 (52% vs 53%; RR 0.98; 95% CI 0.88 to 1.10; P=0.78) and mRS 4–6 (51% vs 53%; RR 0.96; 95% CI 0.85 to 1.08; P=0.48).
  • The trial tested two actively managed temperature targets with fever prevention in both groups; it did not test hypothermia versus “no temperature management”.
  • Temperature separation was greatest early (day 2 highest temperature 36.0 ± 1.5°C vs 37.2 ± 0.7°C; P<0.001), with minimal rebound hyperthermia (>39°C: 0 vs 0) and broadly similar withdrawal patterns (median 5 days in both groups).
  • Serious adverse events were common; hypokalaemia occurred more often with 33°C (19% vs 13%; P=0.018), aligning with expected hypothermia physiology and the trade-offs of deeper cooling.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Overall Takeaway

TTM is a landmark postarrest RCT because it challenged the assumption that 33°C is intrinsically superior to higher controlled temperatures when both arms receive active fever prevention and modern ICU care. Its neutral results redirected the field towards rigorous temperature control (especially avoiding fever) and highlighted methodologic vulnerabilities unique to postarrest trials—particularly open-label co-interventions and withdrawal-of-care pathways—thereby shaping subsequent RCT design and guideline evolution.

Overall Summary

  • 33°C did not improve 180-day survival or functional outcome compared with 36°C in a population receiving active temperature management in both arms.
  • Internal validity was strengthened by concealed allocation, balanced baseline care, high follow-up completeness, and structured withdrawal patterns; the major threat was open-label delivery.
  • The trial reframed “temperature management” as active, protocolised care (including fever prevention), not simply a single hypothermia target.

Bibliography

Last updated January 27th, 2026